Datasets:

joyboseroy
/

imljd

	---
	license: cc-by-4.0
	language:
	- en
	tags:
	- legal
	- india
	- nlp
	- knowledge-graph
	- matrimonial
	- 498a
	- judicial
	size_categories:
	- 1K<n<10K
	task_categories:
	- text-classification
	- token-classification
	- summarization
	configs:
	- config_name: sc
	data_files: sc_enriched.csv
	- config_name: hc_karnataka
	data_files: hc_karnataka.csv
	- config_name: combined
	data_files: hc_matrimonial.csv
	---

	# IMLJD — Indian Matrimonial Litigation Judgment Dataset

	Computational dataset of 4897 Indian court judgments on matrimonial disputes (IPC 498A, DV Act, CrPC 482 quashing petitions), built from AWS Open Data judicial archives.

	## Dataset Description

	Code and knowledge graph: https://gitlab.com/joyboseroy/imljd

	\| Sub-corpus \| Cases \| Court \| Period \| Precision \|
	\|-----------\|-------\|-------\|--------\|-----------\|
	\| SC matrimonial \| 1,474 \| Supreme Court of India \| 2000–2024 \| Medium (broad filter) \|
	\| Karnataka HC \| 2,139 \| Karnataka High Court \| 2018–2024 \| High (482 confirmed) \|
	\| Total \| 3,613 \| \| \| \|
	\| Combined HC \| 3,423 \| Karnataka + Delhi + others \| 2018–2024 \| Mixed \|
	\| Grand Total\| 4,897 \| \| \| \|

	## Key Statistics

	\| Metric \| Value \|
	\|--------\|-------\|
	\| Total cases \| 3,613 \|
	\| SC quash success rate \| 57.6% \|
	\| HC (Karnataka) quash success rate \| 39.7% \|
	\| Cases with CrPC 482 \| 2,179 \|
	\| Cases with IPC 498A \| 192 \|
	\| KG nodes \| 1,520 \|
	\| KG edges \| 13,364 \|

	## Columns

	### SC sub-corpus (`sc_enriched.csv`)

	\| Column \| Description \|
	\|--------\|-------------\|
	\| case_id \| Stable identifier \|
	\| title \| Case title \|
	\| petitioner / respondent \| Party names \|
	\| year \| Year (2000–2024) \|
	\| case_type \| quash / appeal / maintenance / bail / other \|
	\| outcome \| quashed / allowed / dismissed / settled / disposed / partly_allowed \|
	\| statutes \| Pipe-delimited statute list \|
	\| disposal_nature \| Raw disposal string \|
	\| mediation_mentioned \| bool \|
	\| settlement_mentioned \| bool \|
	\| relatives_accused \| bool \|
	\| judicial_criticism_misuse \| bool \|
	\| arnesh_kumar_cited \| bool \|
	\| rajesh_sharma_cited \| bool \|

	### HC Karnataka sub-corpus (`hc_karnataka.csv`)

	\| Column \| Description \|
	\|--------\|-------------\|
	\| title \| Case title (CRL.P/NNNNN/YYYY format) \|
	\| judge \| Presiding judge \|
	\| decision_date \| Date of judgment \|
	\| disposal_nature \| ALLOWED / DISMISSED / DISPOSED / Partly Allowed \|
	\| outcome \| quashed / dismissed / disposed / partly_allowed \|
	\| _year \| Year (2018–2024) \|
	\| _bench \| Bench (karhcdharwad / karhckalaburagi / karnataka_bng_old) \|
	\| statutes \| CrPC 482 (all cases) \|
	\| case_type \| quash (all cases) \|

	## Data Sources

	Both sub-corpora built from AWS Open Data (no credentials needed):
	- `s3://indian-supreme-court-judgments/`
	- `s3://indian-high-court-judgments/`

	## Usage

	```python
	from datasets import load_dataset

	# Supreme Court cases
	sc = load_dataset("joyboseroy/imljd", "sc")

	# Karnataka HC quash petitions
	hc = load_dataset("joyboseroy/imljd", "hc_karnataka")

	# Basic analysis
	import pandas as pd
	df = pd.DataFrame(sc["train"])
	quash = df[df["case_type"] == "quash"]
	print(f"SC quash success rate: {(quash['outcome']=='quashed').mean()*100:.1f}%")
	```

	## Knowledge Graph

	A NetworkX/GEXF knowledge graph is included in the repository:
	- Nodes: Case, Statute, Court, Outcome, Precedent, Year
	- Edges: INVOKES, HEARD_BY, RESULTS_IN, CITES, DECIDED_IN

	Open `data/kg/imljd_graph.gexf` in Gephi for visualisation.

	## Ethical Considerations

	- Public court judgments only
	- Names present as in original public records
	- Recommended: anonymisation pass before NLP model training
	- Not suitable for "false case detection" — ground truth doesn't exist cleanly
	- Framing: procedural fairness research, not case outcome prediction

	## Citation

	```bibtex
	@dataset{boseroy2026imljd,
	title = {IMLJD: Indian Matrimonial Litigation Judgment Dataset},
	author = {Bose, Joy},
	year = {2026},
	url = {https://huggingface.co/datasets/joyboseroy/imljd},
	note = {3,613 cases, Supreme Court 2000-2024 and Karnataka HC 2018-2024}
	}
	```

	Related work:
	```bibtex
	@article{boseroy2026falkor,
	title = {FalkorDB-IRAC: Graph-Grounded Legal Reasoning},
	author = {Bose, Joy},
	year = {2026},
	url = {https://arxiv.org/abs/2605.14665}
	}
	```